An evolutionary approach to crowd simulation

Abstract

In previous work, virtual force has been used to simulate the motions of virtual creatures, such as birds or fish, in a crowd. However, how to set up the virtual forces to achieve desired effects remains empirical. In this work, we propose to use a genetic algorithm to generate an optimal set of weighting parameters for composing virtual forces according to the given environment and desired movement behaviour. A list of measures for composing the fitness function is proposed. We have conducted experiments in simulation for several environments and behaviours, and the results show that compelling examples can be generated with the parameters found automatically in this approach. © 2007 Springer-Verlag Berlin Heidelberg.